How segmented is the Bogota labor market?

How Segmented is the Bogota Labor Market? FILE COPY SWP434 World Bank Staff Working Paper No. 434 October 1980 Prepared by: Gary S. Fields, Consultant Urban and Regional Economics Division Development Economics Department Copyright O 1980 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. The views and interpretations in this document are those of the w- A and should not be attributed to the World Bank, to its affiliated in A organizations, or to any individual acting in their behalf.   The views and interpretations in this document are those of the author and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. WORLD BANK Working Paper No. 434 October 1980 HOW SEGMENTED IS THE BOGOTA LABOR MARKET? The objective of this paper is to clarify the notions of labor market segmentation as they exist in the literature and then to examine the extent to which labor market segmentation can be said to be the cause of income inequality in the Bogota labor market. The paper suggests that, at a minimum, definitions of labor market segmentation should permit identification of who the segmenters are, what the nature of their segmenting actions is and what the effects of these actions are. A review of the literature reveals that few of the used definitions of segmentation measure up to these criteria. The empirical part of the paper first documents the existence of labor market heterogeneity in terms of the usual variables like sex, age, education, migrant status, industry of activity, occupation and location of residence in Bogota. Among male workers in Bogota it is found that workers in different industries do earn different incomes at the same age and education levels, but these differences are not large in magnitude and some differences are not statistically significant. Overall, only a weak correlation appears between income and occupation or industry of employment. Thus if segmentation exists in the sense of different earnings functions for different sets of otherwise equivalent people, only weak evidence is found in Bogota. Prepared by: Gary S. Fields, Consultant Urban and Regional Economics Division Development Economics Department Copyright @ 1980 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A.   PREFACE I am pleased to acknowledge helpful discussions and comments on the first draft from Gregory Ingram, Kyu Sik Lee, and Rakesh Mohan of the World Bank and Jorge Ducci, Walter Galenson, and Olivia Mitchell of Cornell University. The yiews expressed herein are my own and not necessarily those of the World Bank or of Cornell University. This paper is part of a program of research currently being conducted by the World Bank on Bogota and Cali, Colombia. The goal of the program, entitled The City Study, is to increase our understanding of the workings of five major urban sectors - housing, transport, employment location, labor markets and the public sector - in order that the impact of policies and projects can be assessed more accurately. This paper is part of the labor market and income distribution portion of the study which is coordinated by Rakesh Mohan. Other papers in this series are: -Rakesh Mohan "The People of Bogota: Who They Are, What They Earn, Where They Live". World Bank Staff Working Paper No. 390, May 1980. Rakesh Mohan and Nancy Hartline "The Poor of Bogota: Who They Are, What They Do, Where They Live." World Bank: City Study Project Paper No. 11, June 1980.   TABLE OF CONTENTS Page No. I. INTRODUCTION 1 II. THEORIES AND DEFINITICNS OF LABOR MARKET SEGMENTATION 5 A. Criteria for Defining and Establishing Labor Market Segmentation ............ 5 B. Five Suggested Definitions of Segmentation and Associated Tests .................................. 9 C. The Framework for Modeling a Segmented Labor Market 19 D. Econometric Issues ................................ 24 III. STATISTICAL AND ECONOIIETRIC TESTS 31 A. Basic Tabulations and Cross Tabulations .... ....... 33 B. The Single-Equation Non-Interactive Approach ...... 61 C. Inequality Within and Between Groups .... .......... 68 D. Segmentation Schemes .............................. 74 E. Segmentation by Exogenous Income-Determining Factors (Type-i) .................................. 75 F. Segmentation by Endogenous Income-Determining Factors (Type-2) .................................. 79 G. Segmentation by De2endent Variable (Type-3) ....... 87 H. Group Determination and Inter-Group Mobility ...... 92 IV. CONCLUSIONS 95 A. Conceptual Conclusions ............................ 95 B. Empirical Conclusions ............................. 96 C. Needs for Future Research ......................... 98 BIBLIOGRAPHY 100   I. INTRODUCTION The central question confronting development economists as we enter the 1980's is: "Who benefits how much from economic development and why?" In a book now in press (Fields, forthcoming), I try to inform concerned readers both of the lessons of the past and of the questions which remain to be answered. In addition, specifically for the case of Colombia, I have worked for several years to understand in depth what determines incomes and income inequality. Previous works were sum- marized in a paper recently completed for the World Bank (Fields, 1978a). The present paper is yet one more contribution to this line of research. My point of departure is the question: What causes inequality in the distribution of labor market rewards? One answer that is increasingly being offered by analysts at the World Bank and elsewhere is: labor market segmentation.-/ The purpose of this paper is to evaluate the analytical value of the proposition that labor market segmentation causes income inequality in Bogota. Notions of labor market segmentation have a long intellectual history. Mill may have been the first to call attention to labor market imperfections with his analysis of non-competing groups. To Mill, these labor market differences were rooted in capital market differences; without collateral one could not get a loan, and without a loan one could not invest in human capital. Thus, the non-competing groups were seen as resulting from institutional barriers to the ac- cumulation of human capital by the poor. 1/ e.g., Selowsky (1979, p.19) writes: "Two basic trends have prevented improvements in the distribution of incume over time. One is demographic growth unparalleled in most development experiences; the second has been the emergence of strong tendencies in the economy toward dualism and segmentation in most factor markets..." (emphasis added).  -2- Today, we mean something different by labor market segmentation. One definition, though by no means a universally agreed-upon one, is that labor market segmentation exists when workers face different earnings functions depending on their location in the labor market. In a competitive labor market in full equilibrium, workers with identical education and experience would expect equal earnings for equal hours worked. In a segmented market, workers in the less-favored group earn less than similarly-qualified workers in some other group. Why do different earnings functions occur? The standard explanation of segmented markets in less developed countries (LDCs) focuses on the determinants of wage structure. For example the govern- ment may impose different minimum wage policies on firms in the modern and traditional sectors; modern sector firms are more likely to be unionized; and modern industries may pay higher wages to reduce worker turnover. Add to these such factors as discrimination, nepotism and favoritism, public/private sector differentials, foreign-owned/domestically owned differentials, and individual differences in ability, and we see that the possible reasons for different earnings functions are many. There are other problems beyond just the differentials in earnings functions. Why don't workers in the lower earning groups enter the high earning labor markets? Why don't employers who pay high wages hire more workers until the value of the marginal product of labor is equal between groups? The issues then are what determines the size of the various groups, what determines different workers' access to employment and income opportunities, and why barriers to mobility among some groups persist over time. The answers to these questions turn on the nature of the groupings themselves.  -3- Some groupings are based on fixed characteristics. Workers in poor countries cannot choose their sex in order to avoid sex discrimination, nor can they choose to be descendents of conquistadores rather than indios, or have parents who are professionals rather than peasants. In these cases, the determinants of group membership are not at issue; the reasons for earnings differentials are. Other groupings are not predetermined. For example, the number of jobs in various occupations and industries, as well as the access of various groups of workers to those jobs, vary with macroeconomic conditions, hiring practices, and the like. All these aspects of group membership are very much of interest to the following discussion, as are differences in earnings functions among these groups. Part II of this paper formulates the question-how segmented is the Bogota labor market?--more precisely. After establishing criteria for a meaningful definition of segmentation, I evaluate various definitions that have been suggested in the literature, set up an economic model of how personal and employment characteristics inter- relate to determine income in a segmented labor market, and formulate an econometric procedure for estimating these relationships. In writing Part II, I searched for useful approaches in the existing empirical literature on labor market segmentation in developed countries; I reviewed the literature surveys by Gordon (1972), Flanagan (1973), Wachter (1974), Cain (1976), and Jackson, Solomon, et al. (1976), as well as many of the basic sources cited therein. I looked also at the less developed country literature, the two most comprehensive references to which are the works of Kannappan (1977)  -4- and Berry and Sabot (1978).-/ Unfortunately, I was unable to draw much specific guidance from the available literature. I find the proposition that labor market segmentation causes inequality in the U.S. or LDC labor markets to be ill-defined in many existing studies, to have been "proven" with inappropriate evidence, and to be virtually indistinguishable empirically from alternative hypotheses which main- tain that inequality arises from still-unmeasured human capital differences among workers, non-uniform utility functions, or compensat- ing differentials. This is not to say that the labor market, in Bogota or elsewhere, is a single unified place with equal opportunity for all and equal outcomes for those who work in it, but rather that appeals to the existing segmentation literature do not get us very far in understanding the inequality and associated wage structures that exist. Part III then presents the results of an empirical investigation of labor market segmentation in Bogota. I first present basic tab- ulations and cross-tabulations. Then turning to multiple regression analysis, I review existing studies and present new evidence using single-equation regression models. Next I proceed to different types schemata for segmenting the labor market and running separate earnings functions for workers in the different segments. Three segmentation schema are distinguished and treated empirically in what follows: segmentation by exogenous independent variables, segmentation by endogenous independent variables and segmentation by the dependent variable. Part IV summarizes the paper's conclusions and discusses topics for further research. 1/ See also Fields (1978b).  -5- II.. THEORIES AND DEFINITIONS OF LABOR M.ARKET SEGMENTATION A. Criteria for Defining Labor Market Segmentation The purpose of defining and measuring segmentation is to see to what extent the segmentation concept helps explain the distribution of economic rewards. To be fully satisfactory, any definition of labor market segmentation should at a minimum meet the following criteria: 1. The definition should not be equivalent to the phenomena to be exDlained. If we are seeking to explain poverty-and inequality, segmentation cannot be defined as the existence of poverty and inequality. Tautological "explanations" are not very informative. 2. A satisfactory definition of labor market segmentation must distinguish actions bv segmenters which lead to labor market inequality from "justifiable" differences among workers. If persons with the same education and experience are paid more in one industry than another, is this prima facie evidence of discriminatory behavior by employers or other actors in the labor markets ? Or does it reflect unmeasured productivity differentials among individuals, attitudinal differences among groups toward work, or the luck that some people have in getting higher-paying jobs when not enough good jobs are available to go around? These latter influences do not constitute labor market segmentation in most people's minds. Hence: 3. The definition of segmentation should in princiDle permit identification of the segmenter. At minimum, any attempt to invoke segmentation as an explanation for unequal labor market outcomes should distinguish between segmentation which occurs in the labor market from  -6- that which occurs prior to the labor market. While lack of educational opportunities for children may contribute to inequality in their earnings as adults, this cannot rightfully be attributed to labor market segmentation. A complete segmentation theory should thus establish who is doing the segmentirg. The scheme suggested in Becker's (1957) classic treatment of discrimination-by employers, by employees, and by customers--remains equally relevant-a quarter century later. In the development context, a further issue is that the lack of development itself may preclude mobility and cause so- called segmentation. 4. The definition of segmentation should in principle permit identification of how the segmenter effects segmentation. Employers, for example, may discriminate by only hiring persons from a given group. Alternatively, their discrimination may take the form of wage differentials in the "same" job. Either practice might be termed "labor market segmentation." The definition of segmentation should make clear what actions do and do not constittite segmentation. If the aforementioned criteria are adhered to, segmentation analysis can potentially be of great help in explaining inequality and poverty. But these are stringent requirements seldom approached. Consequently, the potential of segmentation analysis far exceeds its realization to date. Segmentation concepts have demonstrated beyond any doubt that labor market conditions are not uniform for different groups in the population. If non-uniformity is all we mean by such statements as: "there is labor market segmentation by sex," then "proof" of segmentation is neither surprising nor analytically helpful. However,  -7- the claim of segmentation by sex implies other stronger meanings beyond mere differences. Consider the statement: "Employers systematically discriminate against women by hiring identically qualified men pre- ferentially." This is both more precise than the assertion that "there is labor market segmentation by sex" and, if it were true, would be interpreted by many as evidence of segmentation. Likewise, if it were shown that "employers systematically discriminate against women by paying them less than they do to comparably qualified men," this would also be seen as evidence of segmentation. In other words, there are many labor market actions like preferential hiring and wage discrimination, any one of which is evidence of-segmentation by most definitions. Schematically, this might be indicated as: Action A or Action B Labor market segmentation exists. or Action C or The literature on segmentation commonly suffers from two errors of logic. For example, wage differences between men and women are consistent with labor market segmentation but segmentation may not be inferred from such evidence. This fallacy may be illustrated schematically as:  -8- Evidence consistent with action A or v Labor market Evidence consistent with action B segmentation exists or Evidence consistent with action C or A more subtle fallacy derives from the vagueness of the claim that "labor markets are segmented." If there are 10 actions that constitute segmentation by a particular definition and if only one of those actions is shown to exist, there is still segmentation; it is not valid, however, to infer that all 10 possible actions in fact occur. This flawed reasoning can be illustrated as: Action A and Action B Labor market segmentation exists and Action C and Alas, the segmentation literature is replete with these very mistakes.  -9- B. Five Suggested Definitions of Segmentation and Associated Tests To define what segmentation is, it may be helpful to discuss what segmentation is not. In the standard textbook model of a non- segmented (i.e., homogeneous) labor market, (Reynolds, 1978, pp. 84-85): 1. The attractiveness of a job is measured by the wage. 2. All job vacancies are filled through the market. 3. The labor force is homogeneous. 4. There are as many jobs available as there are workers available. 1 5., Workers and employers are perfectly informed. 6. Vacancies are filled instantaneously. Thus, supply and demand for labor determine the volume of employment and the wage rate paid. The model assumes that the labor market processes and outcomes are the same for everyone, i.e., that all workers receive the same labor market returns. The simplest definition of labor market segmentation takes wage equality as the point of departure. Hence, wqe find in the literature: Definition (i): Heterogeneity of Outcome. Heterogeneity of outcome is the essential characteristic of many definitions of labor market segmentation. Indeed, heterogeneity of outcome is sometimes the sole defining characteristic in empirical research. According to Freedman (1976), segmentation is easy to document: professionals earn more than manual laborers; better educated workers receive higher incomes than less educated workers;  -10- unionized industries pay a wage premium over non-unionized ones; urban incomes are higher than rural incomes; and men are paid more than women. By the heterogeneity of outcome definition, these observations are prima facie evidence of labor market segmentation. These definitions and this type of evidence are unsatisfactory. One problem with the heterogeneity of outcome definition is that no attempt is made to standardize for possible compositional differences between groups. In the case of educational differences, allowance should be made for the period of time when the better educated individuals were in schooland -ere not receiving income. As for male-female differences, it is desirable to standardize for length, quality, and continuity of labor market experience. Failure to consider heterogeneity of individuals is an important conceptual deficiency in some writings on segmentation. More importantly, however, if the concept of segmentation were only to imply that different groups are rewarded differently in the labor market, there would be little controversy over its existence, since equality of outcome obviously does not obtain in modern economies. However, with such a definition, nothing can be explained: the statement "segmentation explains inequality" is a tautology, since by definition (i), segmentation is inequality. The definition of labor market segmentation as "heterogeneity of outcome" must therefore be rejected. In an attempt to improve upon this definition, some writers have proposed:  -11- Definition (ii). Heterogeneity of Outcome Among "Comparable" Workers as a Function of Group in the Labor Market (E.g., Occupation or Industry). Souza and Tokman (1977), for instance, claim (p. 8): "For segmentation in the labor market to exist, persons with equal abilities ought to receive different incomes depending on the stratum of the productive units in which they work." (Translation mine, emphasis added.) Virtually the same conception is used by Altimir and Pinera (1977). Likewise, Bourguignon (1979, p. 56) regards segmentation as an "imperfection of the labor market or, in other words, the hypo- thesis that wages in the modern sector are above incomes in the traditional sector" for otherwise identical individuals. (Translation mine.). And, Mazumdar and Ahmed (1977) write (p.1): "A rather stringent definition of labor market segmentation is that a difference in earnings can be attributed to 'institutional' factors after we have allowed for variations in measurable human quality factors like education and experience." -/ These authors present empirical tests in their respective studies covering several Latin American cities but excluding Bogota (Souza and Tokman), several Latin American countries including Colombia (Altimir and Pinera), several Colombian cities including Bogota (Bourguignon), and several Malaysian cities (Mazumdar and Ahmed). In each study, the empirical test follows the same form: multiple 1/ Similar definitions of segmentation have been used in the housing market literature. For instance, Schnare and Struyk (1976) regard a housing market as being segmented when the price of an attribute varies with either structural or neighborhood characteristics.  -12- regressions involving "human capital" and "segmentation" variables. After standardizing for measurable human capital factors like education and experience, these authors find that the occupation or industry of employment is associated with wages or incomes. Hence, they conclude that the respective labor markets are segmented to a greater or lesser degree. 1/, 2/ Another kind of empirical test consistent with Definition (ii) appears in the literature. This involves three steps: first stratifying the labor force by a variable thought to segment the labor market, then running separate earnings functions for the two groups, and finally comparing the regression coefficients using an appropriate analysis of variance test. _/ The literature offers innumerable instances of segmented earnings functions based on such alternative segmentation variables as race, sex, region, occupation, and industry.4/ 1/ Bourguignon sees less segmentation in his evidence than do Souza and Tokman and Mazumdar and Ahmed in theirs. In reading these studies one should be careful to note that the criteria for establishing the existence of segmentation differ from one study to the next. 2/ In their analysis of housing market segmentation, Schnare and Struyk (1976) look at a sample of housing units in the Boston metropolitan area and at various sub-samples defined according to the number of rooms in the house, whether the house is located in an inner or outer suburb, and income. They find that there are statistically significant differences in the effects of various attributes on rent depending on the housing market in question. However, they also note that there is little gain in precision (as measured by the standard error of estimate) when the housing market is stratified by the above-mentioned variables. From this, they conclude that the Boston housing market is not particularly segmented, at least across the range of variables with which they deal. 3/ If the earnings function is a single equation, the appropriate test of equality of regression coefficients is the Chow test described in standard econometrics text, e.g., Johnston (1972). If the earnings model is a multi-equation recursive structure and fits the path-analytical modal of sociologists, the test for the system of equations is given by Specht and Warren (1976). 4/ See Fields and Ducci (forthcoming) for a review of this literature for less developed countries as a whole. The Colombian studies are cited below in Part III.  -13- These approaches might be criticized at several different levels. At this point, I will mention just two of them. One argument is an empirical problem. Some critics would contend that the *included variables (years of schooling and age) fail to capture other important human capital characteristics such as quality of schooling, continuity of experience, extent of on-the-job training, and such personal characteristics as intelligence and motivation. Without statistical controls for these other influences, the possibility remains that workers in the better occupations or industries possess superior human capital which is reflected in their earnings. The missing variables argument clearly contains considerable truth but it can be pushed to the point of nonsense. Those human capital theorists who disbelieve segmentation arguments sometimes go so far as to attribute all of the unexplained earnings differentials to these omitted characteristics. That will not do. It is about as appealing as "explaining" differences in consumer behaviour by a specified but unmeasured list of "taste" differences in utility functions. The second objection is fundamental. Take occupation and industry as examples of segmentation variables. If occupation or industry is significantly related to income after controlling for personal characteristics, or if different earnings functions are found in different occupations or industries, segmentation is said to exist. A severe interpretation problem arises: Does the test of segmentation "prove" segmentation? If it is established that "segmentation" exists by Definition (ii), what does it imply about the functioning of labor markets? W4ho are the segmenters? How do they segment the market? Is not the same regression  -14- result consistent with both benign and malevolent interpretations?l/ The observation that seemingly comparable workers earn more in some employment sectors than in others is consistent with discrimination, screening, and other exclusionary practices; it is also consistent with intersectoral differences in unmeasured working conditions, unmeasured differences among workers in productivity-related characteristics, and heterogeneity in workers' preferences. We have a classic identification problem. The "test" of the phenomenon under study is not a sufficient test--it is a necessary test of a particular kind of segmentation. Definition (ii) is framed in terms of a symptom which may or may not reflect an underlying pathology: discriminatory barriers to l/ Here again, the parallel between the labor market and housing market segmentation literatures may offer insights. Just as Schnare and Struyck sought to claim from evidence of different hedonic prices of housing attributes in different markets that the housing market is segmented, many labor market analysts seek to claim that the labor market is segmented insofar as people in different labor force groups receive different gains in income for each additional year of education depending on their occupation or industry. But in Schnare and Struyk's analysis, and in others to which they refer, no attempt was made to explain why it is that people live in housing markets with higher hedonic prices. If, in fact, land is cheaper in Waltham, or if an extra bedroom costs less in Wellesley, why is this? Are there barriers to mobility? Or is the observed configuration an equilibrium one in the sense that people trade off number of rooms for number of acres? Whether the observed pattern can meaningfully be said to reflect segmentation or not depends on why these differences in prices of land and prices of rooms arise. The same holds for labor market segmentation. The critical questions are why there are different wage structures in some occupations or industries as compared with others and why people work in the particular occupations or industries that they do. The mere finding of differences is not sufficient to establish discrimination against some and in favor of others.  -15- entry into the higher-paying occupations or industries. Besides studying differences in rewards among various groups in the labor market, we thus need to examine differences in access to earnings opportunities. This suggests: Definition (iii): Heterogeneity of labor market functioning in various submarkets. Edwards, Reich, and Gordon (1975) write: The labor market consists of those institutions which mediate, affect, or determine the purchase and sale of labor power; the labor process consists of the organization and conditioning of the activity of production itself, i.e., the consumption of labor power by the capitalists. Segmentation occurs when the labor market or labor process is divided into separate sub- markets or subprocesses or segments, distinguished by different characteristics, behavioral rules, and working conditions. (Emphasis in the original) (p. xi) This definition has been used in effect by many writers including dualists such as Doeringer and Piore (1971), Bluestone (1970) and Harrison (1972) and radicals such as Wachtel and Betsey (1972) and Bowles and Gintis (1975). This definition of labor market segmentation has the virtue of focusing on the functioning of labor markets; its limitation is that by itself it does not explain why the submarkets or subprocesses are heterogeneous. Economists suggest many reasons why submarkets might differ: heterogeneity among workers, non-competing groups in the labor force, different non-monetary satisfactions received in different jobs, monopsony elements in the labor market, monopoly elements in the product market, limited and costly information,, limited and costly mobility, and institutional regidities and regulations. Any of these real world deviations from the simple textbook model of labor markets would result in non-uniform lab6r market processes and unequal outcomes.  -16- While such occurrences suggest the existence of labor market segmentation, we must ask why segments differ. Indeed, segmentation theorists would have us believe that labor markets function in particularly restrictive ways, i.e., that some individuals are prevented from entering a preferred occupation, moving to a higher paying location, acquiring further education and training, or in some other way improving their economic position. This suggests another, more specific definition: Definition (iv). Limited access to good jobs. A "good job"might be characterized by security, high wages, safe and pleasant working conditions, and/or opportunities for training and advancement. When good jobs are limited in number, "the crux of any theory of labor market segmentation is the mechanism or institutional barriers which truncate competition by precluding mobility between the various labor market segments" (Flanagan 1973, p. 253). A particularly well-known segmentation theory is the dual labor market approach advanced by Doeringer and Piore (1971). As described by Wachter (1974), the dual labor market model advances four hypotheses: First, it is useful to dichotomize the economy into a primary and secondary sector, Second, the wage and employment mechanisms in the secondary sector are distinct from those in the primary sector. Third, economic mobility between these two sectors Is sharply limited, and hence workers in the secondary sector are essentially trapped there. Finally, the secondary sector is marked by pervasive underemployment because workers who could be trained for skilled jobs at no more than the usual cost are confined to unskilled jobs. (p. 639). The critical question that still remains, however, is what limits mobility from the secondary to the primary sector. Since good jobs are not available for all, they must be rationed. This suggests another possible definition:  -17- Definition (v): Non-random access to the available jobs. This definition is used in effect whenever one looks at the proportions of workers from particular groups (e.g., racial, sex, regional) who work in different kinds of jobs. Definition (v) differs from Definition (iv) in that it is concerned not just with different outcomes but with systematically different opportunities; it also takes as given that good jobs are limited in number. Definition (v) concentrates our attention on the rules by which the limited jobs are rationed. If the rationing is found to be at least partly systematic we may then examine why some groups of workers and not others have access to certain jobs. Even now, I worry about using Definition (v) and calling -the result "labor market segmentation." In an LDC, good jobs are scarce and must be allocated among would-be employees. W4hat if differences in access among groups of workers are purely productivity based? Partly productivity-based? Not productivity-based at all. Should all non-random rationing of good jobs be considered segmentation? We have come to the same identification problem as before: the same phenomenon (non-random job access may result from varying causes, some discriminatory, some not).Regardless of whether we term the outcome segmentation or not, we have reached another researchable question: what labor market practices determine which groups get the available jobs? Taken tgether definitions (ii) and (v) are the most helpful concepts of labor market segmentation yet devised because they  -18- direct our attention toward the actual wage- and employment- determination mechanisms in labor markets. They take the first step toward explaining why intergroup labor market differentials exist by showing that intergroup labor market differentials exist in particular dimensions. This focus on real world labor market functioning, as distinct from knee-jerk applications of stylized textbook models, explains much of the appeal of theories of segmented labor markets. Segmentation theorists address fundamental questions about the operation of the labor market and of the eeonomic system more generally. Why do some persons have better opportunities than others? Why is discrimination in the economic system perpetuated? Why is poverty transmitted across generations? Why do labor movements in many countries accept the legitimacy of the prevailing economic order? These and other root questions about the operation of labor markets have not received much attention among orthodox economists. As Gordon (1972) writes (p. 14): "Orthodox analysis... tended to take market structure for granted and probe the determinants of behaviour within those given structures. Some economists sought to develop economic models which dealt directly with these basic concerns about the relationship between labor market structure and income." This suggests that the heart of the distinction between orthodox theories of labor markets and segmentation theories may well lie in the nature of the questions that they address rather than in the way of conceptualizing the behavior of individuals and firms.  -19- C. The Framework for Modelling a Segmented Labor Market The preceding definitions of labor market segmentation direct our attention to the determinants of income and sector of employment as functions of other individual and environmental characteristics. To estimate the relationship among these variables in Bogota, we require a model of how the labor market might be segmented. Eight alternative models are presented in Table 1. They employ the following notation: Y = Income of the Individual PERSCHAR = A vector of personal characteristics (e.g., education, age, migrant status, sex) JOBCHAR = A vector of job characteristics (e.g., occupation, industry) x = Other exogenous variables E = Error term. The components of the PERSHAR and JOBCHAR vectors may differ in the two stages of the multi-equation models.  -20- TABLE 1. Eight Models of a Secrented Labor Market 1*idel Nutber Model Forn of and Nare Description Model Model 1. Single Ecuation Inome as a linear Y = a + 6 PERSCHAR Structural Estimation., ccmbinaticn of + y JOBCGAR + e. Linear Specificaticn, personal and job Full Sample. characteristics. :bdel 2. Single Ecuaticn Incare as a Y. =-a. + B. PERCCiAR Structural Estimation, linear ccabinatimn C 1 1 Linear Specification, of a subset of + JCELR +